From BARRON'S) By Gene Epstein There was no surprise in the strong employment report for March. On the contrary, it's been long overdue. The headline numbers, released Friday by the Bureau of Labor Statistics, show an unemployment rate basically unchanged at 5.7%, and a jump in nonfarm payroll employment of 308,000. As I've mentioned before in this space, three-month averages prove to be much closer to the ultimate revised figures than one-month changes. And in this case, gains in payroll employment are not nearly as high as 308,000, but solid enough. Partly because January and February have already been revised up, the three-month average increase through March has been 171,000; and for the private sector alone, 161,000. Some analysts are dismissing the numbers as an artifact of a) better weather in March and the end of the supermarket strike in California. But the three-month average already factors in the bad weather of January and February. And even after excluding employment gains in food-and-beverage outlets, the private sector still picked up a monthly average of 150,000. The real significance of this report is that all the dots can now be connected. By summer '03, the economy began to grow fast enough for the joblessness of this recovery to come to an end: With growth in gross-domestic product running at an annualized 6.2% in the second half of last year (and probably better than 5% in the first quarter of this year), it was no longer possible to satisfy the increase in demand without hiring more workers. And perhaps just as important, with that kind of growth, new businesses begin to sprout up, creating new jobs in the process. So employment probably began expanding by late-summer of '03, as virtually every labor-market indicator has confirmed -- including weekly claims for unemployment insurance (down), announced layoffs by large companies (down), the index of help-wanted advertising (up), the employment indexes from the Institute for Supply Management and National Federation of Independent Business (all up), the unemployment rate itself (down), and the employment figure from which the unemployment rate is calculated. Nonfarm payroll employment, based on a monthly survey of the payroll records of about 160,000 businesses and government agencies, had resisted the trend -- that is, until now. Manufacturing employment has been flat over the past three months, after falling for so long; based on prospects for the growth of output, it should be bottoming out. Otherwise, the gains have been fairly widespread -- in professional and business services, wholesale and retail trade, construction, finance, real estate and health care. An increase of 171,000 per month is about right for this stage in the recovery; employers are always behind the curve. But if payroll gains do not run 250,000 per month or more by summer, then this influential indicator will again become the piece of the puzzle that doesn't fit. Nor can we forgive or forget the 62,000 average increase in the last four months of '03. (Over calendar `03, nonfarm payrolls stayed flat.) I still believe these figures will eventually be revised up. As I mentioned last month, the Bureau of Labor Statistics is probably underestimating the increase in employment from new-business formation. The agency itself has said that the "model" it uses for this purpose is bad at turning points. Since then, I've heard from business-people who have told me that in this expansion, they are outsourcing more than ever before to small enterprise for services that they used to do in-house. And no, they do not mean offshoring; they mean outsourcing, which in this case they do domestically. A house architect in upstate New York has told me, for example, that contractors no longer have employees on their payrolls, mainly to avoid paying workers' compensation insurance. If these anecdotes are more than just anecdotes, they indicate that the Bureau may be missing more jobs than ever before from small-business formation. In any case, we won't know until early next year, when the figures are finally "benchmarked" to the unemployment-insurance data, which are supposed to be a very universal count. -- The unemployment-rate decline has been a bone in the throat of the doubters ever since it fell to 5.7% in December. This figure, confirmed for the past three months, is printing at either 5.6% or 5.7%, as it did in March -- down from a range of 6.3% to 6.1% from June through September. So what more do they want? Well, the anonymous reporter or reporters at the Economist (the articles are never signed) declared Friday that the unemployment rate for March was not to be believed because "many Americans, despairing of ever finding a job, had dropped out of the work force altogether in recent months and were therefore not counted as unemployed." This false statement may have been inspired by Princeton economist and New York Times columnist Paul Krugman. According to a recent column of his, the fact that the "unemployment rate has fallen since last summer...[is] entirely the result of people dropping out of the labor force" (March 12). But the labor force has increased since last summer, not decreased. Krugman and his possible protege at the Economist believe otherwise because someone ignored a footnote. This footnote I'm referring to appeared quite prominently below the officially posted time series for the civilian labor force on the Bureau of Labor Statistics Website. It reads, "Data affected by population changes in population controls in January 2000, January 2003 and January 2004." Now, anyone who has actually followed the employment numbers will know immediately what the agency is trying to tell us. The recent figures are no damn good, is what it's trying to tell us, and to find out why, you'll have to dig a little. Go on the Internet and click on www.bls.gov/cps/cpspopsm.pdf -- a Bureau of Labor Statistics (BLS) paper posted March 3, 2004 -- to get the real numbers. By the third sentence of that paper, it's clear that the decline you might find in the officially posted data is a mirage. (The math's a bit too involved to go into here.) The official numbers with the big footnote do show a decline in the labor force since the summer of last year; but the real numbers show an increase. So the fall in the unemployment wasn't caused by "people dropping out of the labor force," because people didn't. Is this a practical joke that might fool any professional? Well, the BLS pulls these jokes all the time. Indeed, I don't know of any Wall Street economist, good or bad, who's not aware of these glitches. In that same column, Krugman falls for yet another BLS joke: "40 percent of the unemployed," he informs us, "have been out of work more than 15 weeks, a 20-year record." Wrong again and for the same embarrassing reason. Someone was looking at the officially-posted data. Yet in this case there was no footnote. If there were, it would have said, in some way, that you can't compare the recent figures for long-term unemployment with the same figures 20 years ago. Or if you did, you had to adjust the recent figures downward. Based on an analysis in a BLS paper of March '95 (at www.bls.gov/ore /abstract/ec/ec950090.htm), Krugman's "40 percent" would have to be lowered to 34.2 percent to make it comparable to the pre-1994 data. And that would not even be a 10-year record, much less a 20-year one. Why so? Well, in January '94, the Bureau of Labor Statistics and Bureau of the Census introduced a redesigned questionnaire for the Current Population Survey (CPS). The CPS is the monthly door-to-door survey of about 60,000 households from which all the "Household Data" are derived, including the data on unemployment and unemployment duration. The old version had been in use since 1967, and was clearly in need of an overhaul. Questions were dropped; others added; and virtually all were reworded. The BLS and Census had already expected the new questionnaire to cause certain tallies to change radically. Indeed, the revised version suddenly found many fewer involuntary part-timers, "discouraged" workers, and unemployed "job leavers." It also found many more of the long-term unemployed. Just how the new questions were suddenly bringing new answers is a fascinating story in itself. In the case of the long-term unemployed, I personally believe the old questions were better. In sum, to make an historical comparison between now and 20 years ago, you either had to adjust the old data upward, or the new data downward. If you want to use these numbers to know what's happening in the real world, you have to know these things.